As proud partners of the Montreal AI Ethics Institute, we've put in place strong ethical principles on data use as we believe that it's not only important to create value, we are committed to creating value in the right way.
Berkindale Analytics' Ethical Data Principles.
Proper data governance protocols and data classifications are paramount to making sure data privacy concerns are met. Technology should be reliable and safe in the context of data processing.
Private data should be encrypted according to best standards and made available through a comprehensive entitlement system. Penetration testing and audits of the development process will be performed.
Data systems and downstream AI outputs should be understandable and have transparent data lineage. Full audit trail of actions performed by the data platforms should be readily available.
Firms should be accountable for how their data is used and the outputs of downstream AI systems. Firms must be aware of the type of data and training sets that will be used to feed machine learning applications.
Technology is for all to use in a fair and inclusive way. However, historical data and training sets may have unfair bias and so we support discrimination-aware data mining (DADM) and fairness, accountability, and transparency in machine learning (FAccTML) communities.
Outputs of data processing and AI algorithms must have explainability. Any decision-aid system must also have a human-in-the-loop (HITL) as a failsafe that can catch and fix a decision system's outputs.